Clinical Trials

❘❙❚■ Chapter 11 | Factorial DesignAs prior information was available on the risk of postoperative nausea andvomiting for one treatment group, in order to quantify this effect appropriatelythere were twice as many people randomized to one arm of the interventioncompared with the other. The trial was powered to allow for interactions betweena maximum of three interventions; however, since no interaction effect waspresent, the trial provided an estimate of the combined effect of interventions inaddition to individual effects. It was also not feasible for one study site torandomize individuals to all six interventions; however, even with a reducednumber of individuals, there were enough individuals for the plannedcomparisons to be viable.What are the advantages of a factorial design?CostThe main advantage of a factorial design is its relative economy: it is possible toevaluate two or more interventions within the same trial at less than the cost oftwo separate trials, and possibly with only a marginal additional cost to a singletrial of one intervention. Rather than omitting treatment comparisons in order toperform a conventional parallel-arm study, or expanding the sample size to amultiple-arm study, it is possible to evaluate multiple treatments within the sametrial using fewer patients than individual comparisons.Sample sizeTake, for example, the previously mentioned trial of multivitamins and vitamin Ain HIV-1-infected women [1]. If the same study had been performed as athree-arm parallel study with the same sample size, N, then a third of theindividuals would have been randomized to receive multivitamins, a third wouldhave received vitamin A, and the remaining third would have received neither. Inthe factorial design used, however, half of the women were randomized to eachtreatment irrespective of the other treatment. Therefore, a three-arm parallel trialhas less power to make comparisons; moreover, to achieve the same power as ina 2 × 2 factorial trial, the three-arm parallel trial would need to randomize 1.5Nwomen. Hence, substantially fewer individuals are needed in a factorial trial thanin a multiple-arm parallel study with the same power.Exploring interaction effectsA second, often-quoted advantage is that factorial designs are useful to crudelyevaluate the combination of interventions. If the aim of the study is to accuratelyquantify the interaction effect, many more individuals are required. Quantifyingthe effect of a combination of treatments is represented by a multiple-arm trialthat tests the treatment combination in a distinct arm. It follows that, in the above110